MULTI-FLGANs: Multi-Distributed Adversarial Networks for Non-IID distribution

TL;DR

This paper introduces MULTI-FLGAN, a novel federated GAN architecture designed to improve stability and performance in non-iid distributed datasets, effectively addressing issues like mode collapse and low-quality image generation.

Contribution

The paper proposes MULTI-FLGAN, a new federated GAN architecture that enhances stability and quality in non-iid settings, outperforming existing FLGAN models.

Findings

MULTI-FLGAN is four times more stable than baseline FLGAN.

It achieves higher inception scores on average across 20 clients.

The architecture effectively mitigates mode collapse in non-iid data.

Abstract

Federated learning is an emerging concept in the domain of distributed machine learning. This concept has enabled GANs to benefit from the rich distributed training data while preserving privacy. However, in a non-iid setting, current federated GAN architectures are unstable, struggling to learn the distinct features and vulnerable to mode collapse. In this paper, we propose a novel architecture MULTI-FLGAN to solve the problem of low-quality images, mode collapse and instability for non-iid datasets. Our results show that MULTI-FLGAN is four times as stable and performant (i.e. high inception score) on average over 20 clients compared to baseline FLGAN.